<- Back to Tasks

[Build 118] Critical decrease in computation speed 118 vs 117

There is a huge computation speed difference between 117 (much faster) and 118 (much much slower).

See attached figures!

Example 1: Build 117: Monte-Carlo random ATR test: time per accepted strategy= 8s

Example 1: Build 118: Monte-Carlo random ATR test: time per accepted strategy= 1min. 55s

In both cases exactly the same strategies, configuration and hardware is used.

Attachments

SQ118 MC random ATR.png

(228.67 KiB)

RETEST Nr8 MC random volatility ATR.cfx

(2.79 KiB)

SQ117 MC random ATR.png

(233.42 KiB)

Votes +2
Project StrategyQuant X
Type Bug
Status Refused
Priority Normal

History

coensio

04.02.2019 11:27

Task created

clonex / Ivan Hudec

04.02.2019 11:41

Voted for this task.

clonex / Ivan Hudec

04.02.2019 11:42

It is ok because in 117 was mc test done by selected timeframe precision. I want to ask developers if ther is no way to use all CPU power duting MC test? My cpu is onlyaroundn 10% when MC teste are running

Mark Fric

20.02.2019 15:48

Status changed from New to Refused

it is as Clonex said - from 118, MC test uses the same precision as main data, that's why it is now slower.

As for CPU usage - it is by design that Retester runs as many strategiy tests in parallel as possible, but Monte Carlo tests for each strategy are not run in parallel.

So to use full CPU you must have at least as many strategies to retest as you have cores.

Karish

25.02.2019 12:22

Voted for this task.

Votes: +2